What can AI do for financial services firms?
At an adviser, wealth manager, lender or fintech, AI takes over the recurring reading, writing and lookup work: summarizing documents, preparing reports, following publications from supervisors and answering general client questions. Decisions on advice, credit and acceptance stay with your employees.
A large part of financial work consists of documents. Annual accounts, prospectuses, terms, client files and regulation. Collecting and ordering that information comes before every judgement, and that preparatory work is suited to AI.
There are two ways to approach it. Your employees use an AI assistant such as ChatGPT or Claude themselves for summaries and drafts. Or you have an AI agent set up: a digital worker that is connected to your systems and prepares tasks on its own. What that is, is explained in what is an AI agent.
Which tasks does an AI agent take over in the financial sector?
An AI agent takes over five kinds of tasks: reading and summarizing documents, preparing client reports, following regulation, completing a credit file and answering general client questions.
Reading and summarizing documents
You supply annual accounts, a prospectus or a contract. The agent extracts the key figures and the main provisions and names the page for each point. It puts two versions of a document side by side and marks the differences.
Needed: the documents in a searchable format. Human check: the analyst reads the source for every figure they take over. A language model can misread an amount or skip a footnote.
How this works is explained in AI agent for document processing.
Preparing client reports and conversation reports
The agent puts the figures from your administration into the format of your periodic report and writes a draft commentary. From a client conversation it produces a draft report. Before a meeting it lists that client's recent changes and open items.
Needed: the client data from your CRM or portfolio system and your own format for reports and file notes. Human check: the adviser checks the figures, adds to the commentary and approves the report. The adviser remains responsible for the file.
Following regulation
The agent reads new publications from, among others, the AFM, DNB and the European supervisory authorities and writes a short summary of each publication with the source attached.
Needed: the list of sources you want to follow and the topics that matter for your licence. Human check: the compliance officer assesses what a publication means for your organization and which action is needed.
Completing a credit file
The agent checks whether all documents of an application are present, copies the data into your own format and prepares an overview for the underwriter.
Needed: the list of required documents per type of application. Human check: the underwriter assesses creditworthiness and decides. If you have an AI system carry out that assessment of a natural person, that system falls into the high-risk category of the EU AI Act.
Answering general client questions
Clients ask which documents they need to supply, what a term in their statement means or how to pass on a change. The agent answers those questions with your own terms and explanatory texts.
Needed: the current terms and your answers to frequently asked questions. Human check: requests for a product recommendation go to an adviser.
If you work in insurance, also read AI for insurance.
What can you do yourself today with ChatGPT or Claude?
With a ChatGPT or Claude subscription you can do four tasks faster yourself today, without integrations:
- Summarize public annual accounts with references to the pages
- Draft a client letter based on your own notes
- Summarize a supervisor's publication for your team
- Turn a meeting into a report with the agreements made
Remove personal data and confidential client information before you enter text, unless you use a business account with a data processing agreement. Check every figure against the source. What is a prompt explains how to write a good instruction.
Which rules apply to AI in the financial sector?
Five frameworks apply to AI in the Dutch financial sector: the GDPR, the Dutch Financial Supervision Act (Wft), the EU AI Act, DORA and supervision by the AFM and DNB. This describes what those rules ask for, as of October 2026. Put your own situation to your compliance officer or a lawyer.
GDPR. Clients' financial data is personal data. Sign a data processing agreement with every AI supplier and give the agent access only to what it needs for the task. Read more in using AI in a GDPR-proof way.
Automated decisions. As a main rule, Article 22 of the GDPR prohibits decisions with legal or similarly far-reaching effects that are made in a fully automated way. Rejecting a credit application is such a decision. Where the law allows an exception, the client has the right to human intervention.
Informing and advising. The Dutch Authority for the Financial Markets (AFM) distinguishes between informing and advising. A personal recommendation for a specific financial product is advice within the meaning of the Wft. According to the AFM, automated advice must meet the same legal requirements as advice given by a person. Configure the agent to explain what the terms say and to pass requests for a recommendation on to an adviser.
EU AI Act. AI systems that evaluate the creditworthiness of natural persons or establish their credit score are listed in Annex III as high-risk. Systems for detecting financial fraud are excluded from that category. Since the amendment of July 2026, the high-risk rules apply from 2 December 2027. The transparency rules have applied since 2 August 2026: a client must know they are communicating with an AI system. The dates and obligations are in the EU AI Act explained.
DORA. The EU regulation on digital operational resilience has applied to financial entities since 17 January 2025. It sets requirements for managing ICT risks and for arrangements with external ICT service providers. Check whether your AI supplier falls under it.
Supervision. In April 2024 the AFM and the Dutch central bank (DNB) published the report The impact of AI on the financial sector and supervision. They expect financial institutions to use AI responsibly and state that existing regulation also applies to AI applications.
| Use | Who decides | Watch out for |
|---|---|---|
| Summarizing a document | The analyst | Read the figures in the source |
| Draft client report | The adviser | Only the client data needed |
| General client questions | The agent informs | State that the client is communicating with AI |
| Product recommendation | The adviser | Advice within the meaning of the Wft |
| Creditworthiness of a person | The underwriter | High-risk under the AI Act |
How do you start with AI at a financial services firm?
You start with one internal task where the agent prepares and an employee decides.
Step 1: choose the task. Summarizing public documents or following regulation are suitable starting points, because no client data is needed for them.
Step 2: involve compliance. Record which data the agent may use, who checks the outcome and which agreements you make with the supplier.
Step 3: test with existing work. Have the agent summarize documents your team has already assessed and compare the outcomes.
Step 4: record the check. Note per task who reviews the result and keep what the agent supplied with the file.
Step 5: expand. After that, add a task with client data, such as conversation reports or draft reports.
Frequently asked questions
Is an AI agent allowed to give financial advice? A personal recommendation for a financial product is advice within the meaning of the Dutch Wft. According to the AFM, automated advice must meet the same legal requirements as advice given by a person. Have an AI agent inform and pass advice questions on to an adviser.
Is AI for credit assessment high-risk under the EU AI Act? Yes. AI systems that evaluate the creditworthiness of natural persons or establish their credit score are listed in Annex III as high-risk, with the exception of systems for fraud detection. The rules apply from 2 December 2027.
Is AI allowed to reject a credit application on its own? As a main rule, Article 22 of the GDPR prohibits fully automated decisions with far-reaching effects on a person. Rejecting a credit application is such a decision. Have an underwriter decide based on the file the agent prepared.
What do the AFM and DNB expect of AI in the financial sector? In their report of April 2024, the AFM and DNB write that they expect financial institutions to use AI responsibly. According to the supervisors, existing regulation also applies to AI applications.
What does AI cost for a financial services firm? A 1-on-1 AI training costs from €449 excl. VAT. The implementation of an AI agent starts from €1,495 one-time, on invoice (as of October 2026). Read more about what an AI agent costs.
Using AI in your organization?
Curious what a digital worker can concretely do for financial services firms? Check our AI agent for financial services firms or read what an AI agent costs. Rather upskill your team first? Our 1-on-1 AI training covers that. Or book a free intro call.
